MiniMax
availableShows if the model has enough results for an index.MiniMax M3
MiniMax M3 is a non-reasoning model from MiniMax. 48 benchmarks count toward its score, in 7 categories.
IndexOverall score. 50 is the middle.56.9 ±2.8
CoverageShare of the index weight with results.95%
SpeedOutput tokens per second.64/s
Input / 1MUS dollars per 1M input tokens.$0.3
Output / 1MUS dollars per 1M output tokens.$1.2
ContextMaximum tokens in one request.1.05M
EloLMArena rating and rank.1433 (#82)
50 is the middle of the board. The range shows the doubt in the index.
48,540 votes. Elo shows what people prefer. It does not change the score.
CapabilitiesScore per category. 50 is the middle.
50 is the middleResults
48 counted| BenchmarkThe test name. | CategoryThe capability that the test measures. | ResultThe score from the publisher. | IndexThis result on the index scale. | RunThe settings of the run. | DateDate of the result. | Published byThe source of the result. |
|---|---|---|---|---|---|---|
| Artificial Analysis GPQA Diamond | Knowledge | 92.9% | 63.8 | — | — | Artificial Analysis |
| GPQA Diamond | Knowledge | 92.7% | 63.8 | — | 1 Sept 2026 | Vals AI |
| OmniDocBench 1.5 | Multimodal | 91.6% | — | — | — | OpenAI |
| τ²-Bench Tool-Agent-User Evaluation | Agentic | 88.9% | 62.6 | — | — | Victor Barres et al. |
| Artificial Analysis Harvey LAB-AA | Agentic | 88.4% | 70.5 | — | — | Artificial Analysis |
| United States of America Mathematical Olympiad 2026 | Math | 85.7% | — | — | — | Mathematical Association of America |
| Video-MME with subtitle | Multimodal | 85.4% | — | — | — | Qwen |
| VideoMMMU | Multimodal | 84.6% | — | — | — | Qwen |
| MMLU Pro | Knowledge | 84.2% | 53.2 | — | 1 Sept 2026 | Vals AI |
| BrowseComp | Agentic | 83.5% | 69.9 | — | — | OpenAI |
| Artificial Analysis Long Context Reasoning | Reasoning | 83.0% | 65.7 | — | — | Artificial Analysis |
| Artificial Analysis IFBench | Instruction | 82.9% | 73.2 | — | — | Artificial Analysis |
| LiveCodeBench | Coding | 82.2% | 59.0 | — | 1 Sept 2026 | Vals AI |
| GPQA diamond | Knowledge | 81.3% | 53.3 | none effort | — | Epoch AI |
| MMMU Pro | Multimodal | 81.2% | 59.4 | — | 1 Sept 2026 | Vals AI |
| Software Engineering Benchmark Verified | Coding | 80.5% | 62.1 | — | — | Carlos E. Jimenez et al. |
| Artificial Analysis MMMU-Pro | Multimodal | 78.6% | 63.1 | — | — | Artificial Analysis |
| Massive Multi-discipline Multimodal Understanding Pro | Multimodal | 78.1% | 54.4 | — | — | MMMU-Pro authors |
| LiveBench Mathematics | Math | 76.9% | 50.6 | — | 25 Jun 2026 | LiveBench |
| LiveBench Language | Knowledge | 76.8% | 63.3 | — | 25 Jun 2026 | LiveBench |
| LiveBench Data Analysis | Reasoning | 76.2% | 61.8 | — | 25 Jun 2026 | LiveBench |
| BankerToolBench | Agentic | 76.1% | — | — | — | MiniMax |
| SWE-bench | Coding | 75.0% | 57.7 | — | 1 Sept 2026 | Vals AI |
| GDPval rubrics | Agentic | 74.7% | — | — | — | MiniMax |
| Claw-Eval | Agentic | 74.5% | 75.5 | — | — | Bowen Ye et al. |
| LiveBench Reasoning | Reasoning | 74.5% | 59.4 | — | 25 Jun 2026 | LiveBench |
| MCP Atlas | Agentic | 74.2% | 64.0 | — | — | OpenAI |
| OSWorld-Verified | Agentic | 70.1% | 58.9 | — | — | Tianbao Xie et al. |
| LiveBench Coding | Coding | 68.2% | 51.0 | — | 25 Jun 2026 | LiveBench |
| SVG-Bench | Coding | 63.7% | — | — | — | MiniMax |
| SWE-bench Pro | Coding | 59.0% | 61.1 | — | — | Xiang Deng et al. |
| Artificial Analysis Coding Index | Coding | 58.6% | 60.2 | — | — | Artificial Analysis |
| LiveBench Instruction Following | Instruction | 57.5% | 51.0 | — | 25 Jun 2026 | LiveBench |
| Terminal-Bench 2.1 | Agentic | 53.6% | 55.7 | — | 21 Sept 2026 | Vals AI |
| SkillsBench | Coding | 51.5% | 66.6 | OpenHands | 11 Sept 2026 | Vals AI |
| VIBE V2 | Coding | 50.1% | — | — | — | MiniMax |
| OpenHarmony Bench v1.0 | Coding | 48.4% | 57.1 | — | — | OpenHarmony Bench authors |
| Vibe Code Bench v1.1 | Coding | 47.6% | 62.0 | OpenHands | 21 Sept 2026 | Vals AI |
| Artificial Analysis SciCode | Coding | 47.1% | 58.2 | — | — | Artificial Analysis |
| Terminal-Bench 2.0 | Agentic | 46.1% | 55.6 | — | 4 Jun 2026 | Vals AI |
| OfficeQA Pro | Multimodal | 45.1% | 57.4 | — | — | OfficeQA Pro authors |
| Terminal-Bench Hard | Agentic | 42.4% | — | — | — | Artificial Analysis |
| NL2Repo | Coding | 42.1% | 58.0 | — | — | MiniMax |
| LiveBench Agentic Coding | Agentic | 40.7% | 55.0 | — | 25 Jun 2026 | LiveBench |
| Artificial Analysis Humanity's Last Exam | Knowledge | 39.0% | 66.8 | — | — | Artificial Analysis |
| GDPval-AA normalized | Agentic | 36.5% | 63.6 | — | — | Artificial Analysis |
| Artificial Analysis EnterpriseOps-Gym | Agentic | 32.1% | 51.4 | — | — | Artificial Analysis |
| Artificial Analysis Agentic Index | Agentic | 30.8% | 62.3 | — | — | Artificial Analysis |
| Artificial Analysis Intelligence Index | Knowledge | 29.2% | 58.9 | — | — | Artificial Analysis |
| KernelBench Hard | Coding | 28.8% | — | — | — | MiniMax |
| OTIS Mock AIME 2024-2025 | Math | 26.7% | 26.5 | none effort | — | Epoch AI |
| Code Migration | Coding | 19.9% | 57.7 | — | 21 Sept 2026 | Vals AI |
| ResearchClawBench | Agentic | 19.8% | — | — | — | InternScience |
| ProofBench v1.1 | Math | 18.0% | 56.7 | — | 21 Sept 2026 | Vals AI |
| Medical Long Context Reasoning (MLCR-AA) | Reasoning | 17.2% | 52.6 | — | — | Wisedocs and Artificial Analysis |
| Artificial Analysis Omniscience Accuracy | Knowledge | 16.7% | 34.5 | — | — | Artificial Analysis |
| Artificial Analysis AnalystAgent | Agentic | 10.0% | 48.5 | — | — | Artificial Analysis |
| Vibe Code Bench 1-100 | Coding | 9.2% | 61.9 | OpenHands | 16 Sept 2026 | Vals AI |
| Mystery Game Puzzles | Reasoning | 8.0% | 42.4 | none effort | — | Epoch AI |
| OSWorld 2.0 | Agentic | 4.6% | 58.2 | — | — | Mengqi Yuan et al. |
| Chess Puzzles | Reasoning | 4.0% | 28.7 | none effort | — | Epoch AI |
| Critical Physics Tasks | Reasoning | 3.7% | 46.1 | — | — | Artificial Analysis |
| Agent Arena command recovery | Agentic | 1.6 | 69.2 | — | 15 Sept 2026 | LMArena |
| Agent Arena steerability | Agentic | -6.6 | 59.9 | — | 15 Sept 2026 | LMArena |
| Agent Arena task outcome | Agentic | -12.0 | 53.8 | — | 15 Sept 2026 | LMArena |
48 benchmarks count, from 54 of 65 results. A grey row does not count. Too few models took that benchmark.